Discussion on newslist relating to our chosen design challenge.

Available as a Word Document here.


Posted September 2nd (pre workshop) by Adrian Bowyer (Engineer)

I like the floating rainbow footballs a lot.

A few technical points:

Avoid running wires to each one if at all possible - they'd be a maintenance nightmare if
the thing is going to be installed for any appreciable length of time.

You could (as you suggest) use wave power.  But it may be simpler (as the tops will be transparent) to just use those sunlight-charged garden lamp systems for storing power,
then junk the actual lamp and replace it with a microcontroller and your LEDs.  There's much less power in sunlight than in waves, but on the other hand the technology can be bought at B&Q...

Maybe use wireless networking to allow everything to talk to everything else.

If you're using a CA to decide the lighting patterns, you may not need a central PC, nor the wireless n/w.  Just have infra-red communication between the spheres, and set the power at a weedy level so that each sphere can only talk to its immediate neighbours.  You'd get interesting effects as the waves sometimes got in the way of the IR, or as a sphere crested a wave and its IR consequently achieved a bigger range from up high.

Get the thing working on terra firma first, then tow it out into the bay...

A land-based one where people walking in-between took the place of the waves interrupting the IR would also be interesting.

Posted October 5th by Mark d’Inverno (Computational Models) and Jane Prophet (Artist)

Thanks to all those who came to the September workshop. We are posting MP3 files from this workshop, and transcriptions in text form in the next few weeks on our web site.  We are also preparing a brief, and partial, overview of the 2 day event which we will post next week.
 
At the end of the workshop we voted on a challenge that we would actually design and produce as a group. To get the ball rolling with discussions about this were posting our summary of the discussion here. Please email your comments to inter-disciplinary@wmin.ac.uk We anticipate a full transcription and an MP3 file being available soon.  The case study chosen was design challenge 1, which you can find at
 
/content/des_ch_summary.html
 
/content/chall_within.html
 
Thanks for all the input on Net Work during the workshop. Here’s a summary of our group, and some one-to-one, discussions about how best to take the project forward. It would be useful now to expand this, and to identify other members’ design challenges that any of this discussion relates to.
 
Adrian Bowyer and Luke Nicholson suggested the following:
To make the floating buoys out of silicon sealant type material (apologies for not remembering the correct name for this material). The buoys would have an air pocket moulded into their design to make them float. All LEDs and electrics could be stuck into the silicon.

Question: how do we address powering up the buoys? Is conductive power (via an electric toothbrush type unit or power pad with no direct wiring) an option?
 
The LEDs would be RGB to enable the widest range of colours to be output.
 
The buoys communicate with each other via light sensors that sense the colour of nearby buoys lighting states (the colour being emitted by nearby buoys). This neatly solves the problem of whether the buoys are controlled by a central computer and makes then collection of buoys act like cellular automata. It also means that if 1 buoy fails it can be changed without too much impact on the whole piece, and that the net will continue to work even with a buoy that is broken.

Question: what type of light sensors do we need? How sensitive are they to RGB values rather than general light levels?
 
During discussion about whether we could make the net work of buoys respond to wave movement, Adrian suggested that magnetic rocker switches would be a cheap solution to detection.
Question: how can we use this meaningfully? If we combine too many sensors will the resulting behaviour of the net work be confusing?
 
People shining torches or other lights onto the floating buoys could influence their behaviour as they would sense these light sources.
Question: what should the net work ‘do’ in response to sensing torch light?
 
Catherine Watling suggested using a radial rather than grid like structure, this is more cell-like and therefore more in keeping with the simulation that inspired the work. She introduced the idea of thalassotherapy, in particular the practice of injecting sea water into people (DATE and URL) and monitoring their physiological changes. Neil Theise commented that sea water has the same salinity as human blood because we evolved from it. Neil also suggested that if the LEDs were very bright, and partly submerged, then the edges of the buoys would be blurred and that this would be an appropriate model for cells, and for cell boundaries ‘disappearing’ (this relates to part of his presentation where he discussed “from bounded cell to body of liquid”).

Question: of we develop a radial structure does this impact on the cellular automata behaviour?

Question: Catherine and Neil will you respond to this and add more info on my mis-representation of your ideas?
 
There was some discussion about whether a similar structure should be sited on land on a promenade as well as the one floating on the sea, with an alternation between the two nets representing the behaviour of people on the promenade to the cellular automata’s behaviour.

Question: what would this add to understanding and enjoyment? Could we implement this more cheaply as a projection from a CA system that was responding to ground-based sensors in the first instance to test it rather than building it with LEDs which would be costly?
 
If the silicon buoys work we could apply this design solution, minus air pocket, to make tennis ball sized ones and use them to make drifts of large numbers of them. This ties into earlier discussions with Rob Saunders about making drifts of LEDs.

Question: Rob will you respond to this?
 
Finally, where should we start in developing a prototype of the floating buoys?

Posted October 6th by Adrian Bowyer (Engineer)

  Adrian Bowyer and Luke Nicholson suggested the following: 
  To make the floating buoys out of silicon sealant type material
  (apologies for not remembering the correct name for this material).

Polydimethylsiloxane, or - if you are a human being - PDMS.  Safety info at:

http://www.inchem.org/documents/icsc/icsc/eics0318.htm

It comes as a liquid with the consistency of golden syrup.  It is colourless and
slightly cloudy, and it sets like that - i.e. almost completely transparent.  To
make it set you mix it 10:1 with polyol (which is basically like a long-chain
alcohol with lots of OH groups stuck on), pour it, and leave it for a day.

It sets faster at higher temperatures.  If it's important to get all the bubbles
out of it when it's set, you can cast it under vacuum.  We have such a vacuum
chamber, but it could only do one football-sized moulding at a time.

  The buoys would have an air pocket moulded into their design to make
  them float. All LEDs and electrics could be stuck into the silicon.
  Question: how do we address powering up the buoys? Is conductive
  power (via an electric toothbrush type unit or power pad with no
  direct wiring) an option?

I know I was rather against running wires to the buoys, but if you run a couple
of power wires in to the circuitry, the PDMS will seal them as it sets.  We do
need to be able to easily swap out each buoy though.

  The LEDs would be RGB to enable the widest range of colours to be
  output.

See:

http://uk.farnell.com/jsp/endecaSearch/partDetail.jsp?SKU=8530203&N=401

for cheap ones (we'd need lots of these per buoy) and

http://uk.farnell.com/jsp/endecaSearch/partDetail.jsp?SKU=8592802&N=401

for ones that aren't.

Note that when all three (R, G and B) are full on in the second one, it's
burning about 5 amps.  To put that in perspective, a car battery is about 30
Amp.hours; i.e if you ran these off a car battery they'd last 6 hours. (Actually
it's a bit better than that, as they only need 7 volts, not 12, so say 8 hours.)

Also 5A  [subsequently changed to 3.5A in a later posting] at 7V is 35 watts, or two soldering irons.  That's some heat to dump.Good thing there's plenty of water about, but we may need to run a tube through the thing to get convective heat transfer.

I can feel an experiment coming on: how bright does each buoy need to be?  Stick one of these in a transparent plastic lunch box with a wire coming out, row out into the bay at night, sling it over the side, and turn it up and down while
someone on the pier with a mobile phone says, "Brighter....   Brighter..."

  The buoys communicate with each other via light sensors that sense
  the colour of nearby buoys lighting states (the colour being emitted 
  by nearby buoys). This neatly solves the problem of whether the buoys
  are controlled by a central computer and makes then collection of
  buoys act like cellular automata. It also means that if 1 buoy fails
  it can be changed without too much impact on the whole piece, and
  that the net will continue to work even with a buoy that is broken.
  Question: what type of light sensors do we need? How sensitive are
  they to RGB values rather than general light levels?

Here's a blue one:

http://uk.farnell.com/jsp/endecaSearch/partDetail.jsp?SKU=4891247&N=401

If you look at page 5 of its datasheet it gives the spectral response of this one and two of its chums.  With a filter these peak at 470 nm (blue), 550 nm
(green), and 620 nm (red).

The LED data are these:

Wavelength, blue LED:475nm; Wavelength, green LED:535nm; Wavelength, red
LED:629nm;

So that's a good match.  But again we'd need to experiment to check how strong a signal could be detected at a range of (say) 1m, and whether we'd need any additional optics to focus the light.

  During discussion about whether we could make the net work of buoys 
  respond to wave movement, Adrian suggested that magnetic rocker
  switches would be a cheap solution to detection.
  Question: how can we use this meaningfully? If we combine too many
  sensors will the resulting behaviour of the net work be confusing?

Good point, which ought to be answerable by simulation, rather than on site.

  People shining torches or other lights onto the floating buoys could
  influence their behaviour as they would sense these light sources.
  Question: what should the net work ‘do’ in response to sensing torch
  light?

I don't think it needs to respond to it any differently to the way the nodes
each respond to their neighbours.   I think the thing to do is to set it up so
that it works autonomously, and it should then respond to external stimuli too.

  Catherine Watling suggested using a radial rather than grid like
  structure, this is more cell-like and therefore more in keeping with
  the simulation that inspired the work. She introduced the idea of
  thalassotherapy, in particular the practice of injecting sea water
  into people (DATE and URL) and monitoring their physiological
  changes. Neil Theise commented that sea water has the same salinity
  as human blood because we evolved from it. Neil also suggested that
  if the LEDs were very bright, and partly submerged, then the edges of
  the buoys would be blurred and that this would be an appropriate
  model for cells, and for cell boundaries ‘disappearing’ (this relates
  to part of his presentation where he discussed “from bounded cell to
  body of liquid”).

  Question: of we develop a radial structure does this impact on the
  cellular automata behaviour?

If each cell still has four neighbours (two radially in and out, two
circumferential), no.  It's just a mapping from cartesian to polar.  But if the
ones near the middle are a sensible distance apart, then the outer-edge ones
will be too far apart, so extra cells will have to be added with increasing
radius to keep the spatial density roughly constant.

  Finally, where should we start in developing a prototype of the
  floating buoys?

I could do some initial dry-land experiments on sensitivity of detectors etc.  I
could also help with PDMS moulding.

Incidentally, PDMS costs about £200 for 25 kilos.  It has roughly the density of
water so a kilo is about a litre.  A football with a diameter of 0.2m has a
volume of 4.2 l, so (ignoring buoyancy bubbles) would use £35 of PDMS, which
seems OK.

Posted October 6th By Andy Adamatzky (Cellular Automata)

I do not really think radial structure will be good, so behaviour will be dull. I would suggest either rectangular or hexagonal lattice (CA with 8 or 6 neighbours). Also, just to be sure --

Light input == real valued
LED output = discrete valued (RGB), almost real valued then :)
right?

Depending on circuit inside each cell we can either use very simple threshold cut-off of inputs (e.g. >0.5 activates cells), or use more compleciated polynomial functions (so several values of light will activate the cell, e.g. 0.2, 0.5, 0.8, to get more "nonlinear" behaviour).

Is there any chance to make them re-programmable, i.e. by switching off and on power supply we cause chips inside cells to cycle through set of cell state transition funcions?

Also, to make patterns persistent it would be better to make boundaries periodic.

May be chips in cells can communicate with each other using bluetooth ?

Pardon - more questions then suggestions.. May be we could make an interactive model of this stuff in Processing, so all people can experiment and choose their functions, and then we can select best solutions?

Posted October 6th by  Paul Brown (Artist)

  Light input == real valued
  LED output = discrete valued (RGB), almost real valued then :)
  right?

You can attenuate LEDs by strobing them (this also has the added value
of cutting down on power consumption).  However in my experience this
causes lots of RF interference (especially in arrays like net work) and
this will be a consideration in a public space where RF is legislated.

Posted October 6th by Andy Adamatzky (Celluar Automata)

  You can attenuate LEDs by strobing them (this also has the added value
  of cutting down on power consumption).  However in my experience this
  causes lots of RF interference (especially in arrays like net work) and
  this will be a consideration in a public space where RF is legislated.

I agree pulse-width modulation (strobing) is the best way of controlling
brightness, and you can virtually eliminate RF problems by doing it at a low
frequency.  A few kilohertz cannot be resolved by the eye and so appears
continuous, but go that low and even the harmonics are too slow to cause RF
problems.

Another advantage might be that most microcontroller chips have inbuilt PWM
ability, though if we need three channels (RGB) that probably won't be a lot of
use as most chips only have one or two.  However, PWM is trivial to implement in
software.

Posted by October 6th byAdrian Bowyer (engineer)

  I do not really think radial structure will be good, so behaviour
  will be dull. I would suggest either rectangular or hexagonal
  lattice (CA with 8 or 6 neighbours). Also, just to be
  sure --

This is something else that should be easy to experiment with using simulation.

  Light input == real valued
  LED output = discrete valued (RGB), almost real valued then :)
  right?

  Depending on circuit inside each cell we can either use very simple
  threshold cut-off of inputs (e.g. >0.5 activates cells), or use
  more compleciated polynomial functions (so several
  values of light will activate the cell, e.g. 0.2, 0.5, 0.8, to get
  more "nonlinear" behaviour).

If each cell has a microcontroller the outputs will be RGB brightnesses probably coded as PWM (see previous e-mails from Paul and me).  That PWM will be picked up by the detectors (which, unlike our eyes, will certainly have the frequency response to discriminate the individual pulses).  Some microcontrollers can decode PWM directly (the PIC16F628, for example), but in the absence of that a resistor and capacitor will integrate PWM to a voltage level very simply, and that can be read using an  analogue-to-digital converter (sticking, for no good reason, with the PIC16F628 - that has an internal voltage reference and comparator that makes a crude [4 bit] a-to-d).  You then have a number representing brightness that the internal program can do what it likes with.

  Is there any chance to make them re-programmable, i.e. by switching
  off and on power supply we cause chips inside cells to cycle
  through set of cell state transition functions?

That would certainly be possible if we used eeprom microcontrollers that can save their state through a power-off.  Lots can.

  Also, to make patterns persistent it would be better to make
  boundaries periodic.

Another alternative would be to make the non-linear function in the cells tend to switch on lights when surrounded by darkness, and switch them off when surrounded by light.  However, that's effectively negative feedback, and the system might tend to settle into a dull non-changing state.

Incidentally, this shouldn't be a serious problem, but if we use light for signalling, we have to prevent cells being able to see themselves, or they'll get swamped by their own light.  Remember there'll be a lot of light backscatter by the surrounding seawater.

  Maybe chips in cells can communicate with each other using
  bluetooth?

Such a scheme would be more reliable than the light communication.  The light
idea is beautiful because it uses the very essence of the work as its medium of
internal signalling, and also it so easily allows public interaction.

But we mustn't get too seduced by elegance into a design that does not have
engineering robustness.  As Jane emphasised: it has got to work, and work very reliably.

Posted by Stasha Lauria October 6th (Intelligent Systems)

  I do not really think radial structure will be good, so behaviour 
  will be dull. I would suggest either rectangular or hexagonal 
  lattice (CA with 8 or 6 neighbours). Also, just to be
  sure --

Is it really necessary to adopt a 'regular' structure?

Afterall, if the 'real world' were mainly dominated by such 'high levels' regular patterns, it would not be difficult to describe it with few equations.

Maybe the differences between mechanical and computational, between the
physical and the virtual are in the intrinsic deterministic properties of algorithms (almost by definition, no algorithmic process can possibly generate random numbers).

So the ability to exhibit more irregular properties (at 'high level')  could be the base of a move from closed, discrete worlds to open ones.

Maybe the regularities need to be pushed down at a more lower
('microscopic') level (microcontoller level)?

Posted October 6th By Andy Adamatzky (Cellular Automata)

If elementary units will communicate via bluetooth (not sure about light??) then underlying architecture may be a fully-connected network, then any other kind of regular or irregular architecture can be embedded into such a network. So, yes, unless some engineering limits are imposed, it is better to have fully-connected communication network.

However, we also must always keep in mind that the proposed network is REALLY small to exhibit REALLY attractive spatio-temporal dynamics, take e.g. Game of Life automaton and make lattice size 10x10 -- what you will see? – almost nothing because in all CA models (and even in numerical models of reaction-diffusion equations or other nonlinear systems), wave-length of growing patterns is 2-3 (!) nodes, and mobile self-localisations do usually occupy between 2x2 to 5x5 nodes. So on such small network no one will see cool waves, fractal patterns, Turing patterns or whatever, it will be just flicking of lights.

Posted October 6th By Rob Saunders (Multi-agent Systems)

I've whipped up a really simple simulation of a network.

  http://homepage.mac.com/rob.saunders/lightbouys_01/index.html

The update rule for the lights is just "well-dressed noise" (random noise with a averaging of local light values) but it might give people some ideas of what patterns are possible on such a small grid.

Posted October 6th by Catherine Watling (Artist)

I agree with Stasha

When i suggested a radial structure I was thinking in terms of the piece looking more cellular and responding in a more natural and organic way to it's surroundings.  In terms of drawing parallels to natural cellular function in the body as opposed to making a purely engineered, man-made form it could be good to look at quite highly complex natural structures that appear on a cellular scale.

I had radiolarians and dinoflagellates, plankton and algae in mind when suggesting this.  These are sea dwelling single celled organisms which are incredibly complex.  Also, the fact that they are sea dwelling makes them idea for the context.  These are some radiolarians.  When I first suggested the radial structure the circular radiolarian in the second image was the one I had in mind.

Take a look at some images for inspiration: [to view Catherine's images click here] 

and have a look at these links
http://www.radiolaria.org/index.htm
http://www.nhm.ac.uk/nature-online/virtual-wonders/vrblaschka1.html

does it have to be a 2d plane?

does the grid only have to have 10x10 buoys?
could there be a more complex radial structure of smaller bouys?  that would allow for a more complex CA?

Just some food for thought
Catherine

Posted October 11th by Neil Theise (Stem Cell Biologist)

  Question: Catherine and Neil will you respond to this and add more info on my
  mis - representation of your ideas?

Catherine Watling introduced the idea of  thalassotherapy, in particular the practice of injecting sea water  into people (DATE and URL) and monitoring their physiological changes.....  she went over this with me after the discussion and its interesting but not particularly relevent i think - essentially this thalassotherapy was/is a way of getting "saline solution" for a patient that has ALL the minerals in the serum, rather than just a few, and uses sea water to do that because its so completely mineralized with all sorts of things...  just a reflection of how our body composition is like sea water (of 40 million years ago or whatever)
 
as for the boundary blurring issue, that i think is a very precise analogy for what i've been talking about and if its the light being sensed then its quite like your display of sdf-1 diffusing (green) from your cells in your model display...  it is both light ad a cloud of the signalling molecule communcating to other cells, and more intense for the nearby cells less intense farther away as it diffuses.   so light = cytokines
 
AND in daylight the buoys look like discrete units floating in water (like blood or other tissues), but at night the hard edge of the buoy dissolves into bright light that diffuses toward other buoy's next door, so the buoy-as-thing becomes the daytime metaphor for the body (and perhaps could signal by blue tooth during the day?) and when it gets dark the net becomes a diffuse light thing and the buoy's 'disappear' into a 'fluid' field of shifting light (now signalled by light perhaps?)
 
so the light signalling is an important part of the metaphorical/analogical aspect of the piece.   though of course 'radio waves' could do the same, but they lack the visual part of conveying the message

Posted October 11th by Adrian Bowyer (Engineer)

  Neil Theise: AND in daylight the buoys look like discrete units floating in water
  (like blood or other tissues), but at night the hard edge of the buoy dissolves into
  bright light that diffuses toward other buoy's next door, so the buoy-as-thing
  becomes the daytime metaphor for the body (and perhaps could signal by blue
  tooth during the day?) and when it gets dark the net becomes a diffuse light thing
  and the buoy's 'disappear' into a 'fluid' field of shifting light (now signalled by
  light perhaps?)

Of course if the individual buoys are inhibited by light and stimulated by darkness as a part of their more complex behaviour, then the whole installation rather neatly turns itself off at dawn and on again at dusk.

Would you like me to put together the electronics for a single cell as an experiment?  This would be a microcontroller and a single multi-coloured LED with a short C program loaded into it, all under the control of a PC.  The final design might well be different; it's just that that's something I could do quite quickly.

Posted October 11th by Jane Prophet (Artist)

[In response to Rob Saunders Simulation above]

I find this simulation very useful. If we make a prototype buoy, or small network of buoys, and cobine these objects with a simulation that utilises the full RGB spectrum (as opposed to my simple 3 colour CA that I showed during my talk) then I think we have a better chnace of getting a large scale art grant to make the piece in total next year.

[In response to Bluetooth Issues]

I agree that using Bluetooth is less 'elegant' in terms of the original
concept of the buoys emitting light, and being responsive to light. However,
it is most important that the project works consistently and it sounds from
your email that Bluetooth would be more robust and reliable.
Who has experience of using Bluetooth in this way? Any hints and tips?

Posted October 21st by Fredrik Olofsson (Generative Music)

here a suggestion on how to map activity on the pier to the cellular automata net in the water.  how this will work is still a bit vague, isn't it?  or did we agree on keeping it non-interactive?  anyway this is nothing fancy - rather simple and direct i hope.

let's say we set up 64 digital sensors (perhaps ultrasound, infrared or just plates to step on) along the pier - one for each buoy.  they could be distributed to clump in the middle like this...

I     I    I   I  I I IIIII I I  I   I    I     I

and we'd let the 2 middle sensors act as a reset.

the system would work in two modes - either setting the cells or running the CA.  the reset sensors would make the system go into setting_mode and a timer would take it back into running_mode.  so people passing by triggering the reset would force the current running CA to stop what it's working on and set all cells to 0.  then the system would wait for x seconds to allow for more sensor input, lighting up corresponding buoy cells as they are triggered, and after that start to run the CA again with those new settings.  this way visitors could clearly see the effects of initial conditions and, hopefully, start to collaborate and build interesting initial patterns.

the sensors could be mapped to cells like this (pardon ugly ascii drawing).
only reset sensors triggered...

-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -
-  -  -  x  -  -  -  -
-  -  -  -  x  -  -  -
-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -

reset sensors + 2 neighbour sensors...

-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -
-  -  -  x  x  -  -  -
-  -  -  x  x  -  -  -
-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -

reset sensors + 7 sensors towards the beach...

-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -
-  -  -  x  -  -  -  -
x  x  x  x  x  -  -  -
-  -  -  -  -  x  x  x
-  -  -  -  -  -  -  -
-  -  -  -  -  -  -  -

with some planning people could build gliders, acorns, gosper guns etc. and see them evolve down in the water.  the actual CA could be anything, but why not a ordinary game-of-life?  perhaps with a different rule for each day of the week?  monday rule 23/3, tues 23/36, wedn 34578/3678 and so on.

the timer (x seconds) would be set to the time it takes to run from the middle out to one of the end sensors.  (ie kids or people on bikes could light up the whole thing if they put some effort into it).  i'm afraid it'd require some additional electronics on the pier for dealing with sensor input (reset, timer, filtering noise) so it's not as elegant as the flashlight idea in that way.  but not more than a microcontroller in a tiny box could handle i guess.

so, do you thing it'd be fun to play with a setup like this?  better ideas?  of course it'd also work in combo with the (more disruptive) flashlights idea.

(and if i'm totally incoherent please say.  i could try to code a little demo of it or something)

Posted October 6th By Andy Adamatzky (Cellular Automata)

Cool, Frederik,

if we are restricted by two states per buoy then we can also "hardwire" generalized Game of Live:

S=sum of 1s in neighbourhood

0 --> 1 if S in [A,B]
1 --> 1 if S in [C,D]

1 <= A <=B <= 8
1<=C <= D <= 8

or if we can have 3 states, then

0 --> 1 if S in [A,B]
1-- > 2 unconditionally
2 -->0 uconditionally

so 0 = resting, 1 = exsited, 2 = refractory,
variety of cool wave-pattern can be obtained,
e.g. [A,B]=[1,7]  classical waves of excitation, spiral and target waves
     [A,B] =[2,2]  gliders, who produce funny patterns when collide

if using bluetooth then no restriction on topology of neighbourhood, then hexagonal three-state CA would be also OK

three-state CA will be enough to simulate almost all phenomena -- excitation, population dynamics, morphogenesis, diffusion-limited aggregation

Posted October 24th by Adrian Bowyer (engineer)

Attached [below] is a (rather unimpressive) picture of the minimal electronics for a buoy.  It's a PIC16F628 microcontroller (£2) driving a single RGB LED (£5) via
the four resistors (4x10p) you can see (there are two blues in the LED package
to increase power at short wavelengths).  The black round thing is a smoothing
capacitor.  The ribbon cable going off at the top is the power supply and data
communications (i.e. there's no fancy Bluetooth stuff; we're wired here...).
The tape round the LED is to diffuse the emmitted light.

The PIC can generate any colour you like by driving the LEDs using pulse-width
modulation.  Over a repeating cycle time of about 100 microseconds it lights
each LED for a fraction of that time.  Different fractions give different perceived colours and intensities; these look completely smooth and steady, despite the fact that they're actually flashing at about 10 KHz (which is far
too fast for the eye to resolve, of course).  The data wires connect to a PC,
which simply tells the PIC what colour to display.

However, the PIC's program memory is only about 10% used by this, which leaves plenty of space for a pretty fancy on-board CA, which would allow the PC to be junked.  (I just connected it for testing.)

A real system would need to drive a stack of LEDs, not just one.  This would
mean one extra power transistor (50p) for each colour.

Posted October 24th by Jane Prophet (Artist)

Thanks Adrian, it's great to see a physical artefact emerging from our discussions